Architecture maps for when the code writes itself

What it does

This repository is a curated, bilingual knowledge base of 26 tutorials and 25 architecture templates that treats system design as the primary deliverable and source code as mere implementation detail. It covers everything from classic distributed systems and payment flows to AI-native stacks like RAG pipelines and agentic workflows, all without prescribing a specific language or framework. The underlying thesis is that AI-generated code is making syntax fluency obsolete, so the remaining human value lies in mapping constraints, tradeoffs, and failure modes before anything is built.

The interesting bit

The project essentially wagers that the architect will be the last human standing in the software loop. Its curriculum runs from basic C4 diagrams and consistency models to governance strategies for vibe coding and eval-driven CI gates for non-deterministic LLM outputs, with a companion architecture-copilot skill that feeds these constraints directly into Claude Code or Cursor.

Key highlights

• 26 tutorials split into fundamentals, distributed-systems deep dives, practical design exercises, and a final “AI collaboration” section on spec-first versus prototype-first workflows.
• 25 templates spanning classic domains—payments, ride-hailing, collaborative docs, video streaming—and LLM-era systems such as AI gateways and RAG knowledge bases.
• Every template links to real open-source projects and engineering documents for follow-up reading.
• Fully bilingual Chinese-English content with an interactive online reader.
• A companion skill project that turns the decision framework into interactive guardrails inside IDE copilots.

Verdict

Grab it if you are leveling up from implementer to tech lead or prepping for system-design interviews; ignore it if you want copy-paste code or framework-specific tutorials.